Hi, iova alloc/free causes big lock contention, which could be easily demonstrated with iperf workload. Previously I posted a patchset: http://lists.linuxfoundation.org/pipermail/iommu/2015-November/014984.html
the concern is it's not generic. This is another try for the issue. This version implements a per-cpu iova cache for small size DMA allocation (<= 64k), which should be generic enough and so we can do batch allocation. iova free could be easily batched too. With batch alloc/free, iova lock contention disappears. Performance result with this patchset is nearly the same as the previous one in the same test. After this patchset, async_umap_flush_lock becomes the hotest lock in intel-iommu, but not very bad. That will be something we need work on in the future. Thanks, Shaohua Shaohua Li (4): IOMMU: add a percpu cache for iova allocation iommu: free_iova doesn't need lock twice intel-iommu: remove find_iova in unmap path intel-iommu: do batch iova free drivers/iommu/intel-iommu.c | 77 +++++++++++++------ drivers/iommu/iova.c | 177 +++++++++++++++++++++++++++++++++++--------- include/linux/iova.h | 10 +++ 3 files changed, 208 insertions(+), 56 deletions(-) -- 2.4.6 _______________________________________________ iommu mailing list iommu@lists.linux-foundation.org https://lists.linuxfoundation.org/mailman/listinfo/iommu
